vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Merge branch 'master' into gg/llama-kv-cache

Browse source 
ggml-ci
This commit is contained in:
Georgi Gerganov 2025-01-31 15:11:02 +02:00
commit 5d3491e789
No known key found for this signature in database
GPG key ID: 449E073F9DC10735
53 changed files with 4034 additions and 208 deletions
Download patch file Download diff file Expand all files Collapse all files

12
include/llama.h
View file

@ -1259,6 +1259,18 @@ extern "C" {
const char * grammar_str,
const char * grammar_root);
/// @details Lazy grammar sampler, introduced in https://github.com/ggerganov/llama.cpp/pull/9639
/// @param trigger_words A list of words that will trigger the grammar sampler. This may be updated to a loose regex syntax (w/ ^) in a near future.
/// @param trigger_tokens A list of tokens that will trigger the grammar sampler.
LLAMA_API struct llama_sampler * llama_sampler_init_grammar_lazy(
const struct llama_vocab * vocab,
const char * grammar_str,
const char * grammar_root,
const char ** trigger_words,
size_t num_trigger_words,
const llama_token * trigger_tokens,
size_t num_trigger_tokens);
/// NOTE: Avoid using on the full vocabulary as searching for repeated tokens can become slow. For example, apply top-k or top-p sampling first.
LLAMA_API struct llama_sampler * llama_sampler_init_penalties(
int32_t penalty_last_n, // last n tokens to penalize (0 = disable penalty, -1 = context size)